31 Mar
2001
31 Mar
'01
12:26 a.m.
Can RA81/82 disks be reformatted in the field? If so, is this
done through XXDP or is there a way to do it from the disk's
serial port?
My RA82 occasionally lights its FAULT light and extinguishes its
ready light during write operations. The likelyhood of it
faulting appears (but I am not certain) to be correlated with a specific
region of the disk.
About five-ten seconds after the fault, the FAULT indicator turns
off and ready comes back on. At no time is an I/O error generated
that the application or the operating system (2.11BSD) see
so this fault appears to be transient and is resolved by either
the drive or the drive/controller together.
It "smells" like a data write error that's resolved after a few
automatic retries. I'm hoping a low-level reformat could clear it
up. Am I wacked out?
If there's an XXDP exerciser/formatter available I'd appreciate it
if someone could point me there. It's been nearly twenty years since
the last time I even tried running XXDP
Thanks as always,
greg
p.s. The RA81 drive on the same controller never gives any
trouble (yet). I've tried some more basic things like
swapping SDI cables and drive ports but it doesn't make
a difference.
Gregory Travis
Cornerstone Information Systems ATS
greg(a)ciswired.com
812 330 4361 ext. 18
31 Mar
31 Mar
4:43 a.m.
New subject: SparcServer 1000
Greetings,
Is anyone here familiar with the Sun SparcServer 1000 systems? This
morning, I picked one up at a hamfest; it has two CPU boards in it,
each, it appears, has 4 CPUs. I was able to identify one of the
boards from it's part number from what I found on the net; appears to
have 20MHz CPUs; the other board is a higher number, but I've found
little data for it this far. The CPU boards are: 501-2336 and
501-2541. Does anone have any data on these?
Being unfmamiliar with multiprocessor systems, I'm wondering how fast
this will be in comparison to my 170MHz Ultra-1; slower for some
things and faster for others, I guess, while consuming much more
power---is this correct?
Wow, this is one heavy machine! In order to figure out if Solaris 8
can run on it, do I count the total amount of memory in the system, or
the amount of memory per CPU? Haven't identified the memory in it
yet, but one CPU board has 12 banks filled.
--
Copyright (C) 2001 R. D. Davis Ancient guru who see future once say:
All Rights Reserved "wise man who meditate upon immortal wisdom
rdd(a)perqlogic.com 410-744-4900 of Bart Simpson's words: "Don't have a cow,
http://www.perqlogic.com/rdd man" will enjoy interesting life."
10:28 a.m.
New subject: SparcServer 1000
On Sat, 31 Mar 2001, R. D. Davis wrote:
Is anyone here familiar with the Sun SparcServer 1000
systems? This
morning, I picked one up at a hamfest; it has two CPU boards in it,
each, it appears, has 4 CPUs. I was able to identify one of the
boards from it's part number from what I found on the net; appears to
have 20MHz CPUs; the other board is a higher number, but I've found
little data for it this far. The CPU boards are: 501-2336 and
501-2541. Does anone have any data on these?
Those are the XDBus boards, identifying which won't tell you much. The
processor modules plug into the MBus. They're likely to be dual 50 MHz
modules, either with 1 MB secondary cache, or without. Find the part
number for the MBus modules and we can tell you more.
What IS 20 MHz, though, is the SBus on those boards.
The 501-2336 board is an FRU, while the 501-2541 came with the system and
had a pair of SM51 (50 MHz, 1 MB L2 cache) processor modules on the MBus.
Being unfmamiliar with multiprocessor systems, I'm
wondering how fast
this will be in comparison to my 170MHz Ultra-1; slower for some
things and faster for others, I guess, while consuming much more
power---is this correct?
Nah, your U1 is likely to be a whole lot faster in nearly every respect,
especially if it's the Enterprise 170, with its 100 Mbit ethernet and
fast/wide SCSI. The SS1000 will continue to perform under a much heavier
load, though, if it does indeed have 8 CPUs. Same caveats with any SMP
system. The SM51 isn't a slouch, but it is old and comparatively slow.
Wow, this is one heavy machine! In order to figure
out if Solaris 8
can run on it, do I count the total amount of memory in the system, or
the amount of memory per CPU? Haven't identified the memory in it
yet, but one CPU board has 12 banks filled.
Total memory in the system and is installed in matched sets of four
modules. Traditionally, modules for this system were available in 8 and 32
MB sizes.
The 8 MB SIMMs have part number 501-1817 while the 32 MB variety have part
number 501-2196.
Solaris support for the sun4d first appeared in Solaris 2.2 and is
supported in Solaris 8, but "will be dropped any day now".
I think this machine is off topic for the list, though, by about two
years. (;
ok
r.
1:49 p.m.
New subject: SparcServer 1000
On Sat, 31 Mar 2001, r. 'bear' stricklin wrote:
Those are the XDBus boards, identifying which
won't tell you much. The
processor modules plug into the MBus. They're likely to be dual 50 MHz
modules, either with 1 MB secondary cache, or without. Find the part
number for the MBus modules and we can tell you more.
Thanks for the help with this!
Hmmm, it now appears that I have merely a three CPU system, not
an eight CPU system. :-( Ah well. Here's some more info:
- One of the FRUs has what I'm guessing to be a single older CPU
module with a TMX390Z50GF (SuperSparc CPU?) and a TMX390Z55GF, which
I'm guessing to be the SuperSparc cache controller---guessing by
what's marked on that IC's heat sink.
- The other FRU has what appears to be newer CPU modules, and these
modules use a single larger heat sink for both of the larger ICs. In
addition, each of these modules has a fuse and a newer style
xtal. osc., also apparently 50MHz. The numbers on the ICs are
different as well, each having a TMX390Z52GF (SuperSparc CPU?) and a
TMX390Z56GF (cache controller?).
None of the above modules have what appaars to be a typical Sun part
number; I noticed "2617-01 R5V54" marked on a sticker on one of the
modules.
Also found four 1.2 GB hard drives, two of which will be used up by
Solaris.
Nah, your U1 is likely to be a whole lot faster in
nearly every respect,
especially if it's the Enterprise 170, with its 100 Mbit ethernet and
Rats... the U1 is not nearly as interesting looking compared to the SS
1000, and it's so much smaller... and only a single CPU. :-) Of
course, it's surely faster than my 4/110. As this has four 1.2 GB
disks in it, I guess it would be reasonable replacement for the 4/110
and it's stack of four shoeboxes with hard drives... although I really
do hate to retire the 4/110. Perhaps I should just trim the 4/110
system back to one hard drive, tape drive and CD-ROM drive, and then
revive my 3/60 using a then spare shoebox. :-)
fast/wide SCSI. The SS1000 will continue to perform
under a much heavier
load, though, if it does indeed have 8 CPUs. Same caveats with any SMP
system. The SM51 isn't a slouch, but it is old and comparatively slow.
Alas, only three CPUs it seems.
Total memory in the system and is installed in matched
sets of four
modules. Traditionally, modules for this system were available in 8 and 32
MB sizes.
Yeeeehaw!!! I've got the 32MB variety! :-) Well, now it appears that
this machine has at least a fair amount of memory, as there are 12 of
the 32MB modules on each FRU, which amounts to a total of 768 MB of
total RAM... with that much RAM, is it still standard to use double
the amount of RAM for swap as a rule of thumb?
Solaris support for the sun4d first appeared in
Solaris 2.2 and is
supported in Solaris 8, but "will be dropped any day now".
Well, as long as I can install Solaris 8 for now, I'll be happy.
I think this machine is off topic for the list,
though, by about two
years. (;
Oops... I guess this thread needs to be put on hold for a couple of
years. :-) ...unless, of course, my mention of the 4/110 and 3/60
count. :-)
--
Copyright (C) 2001 R. D. Davis Ancient guru who see future once say:
All Rights Reserved "wise man who meditate upon immortal wisdom
rdd(a)perqlogic.com 410-744-4900 of Bart Simpson's words: "Don't have a cow,
http://www.perqlogic.com/rdd man" will enjoy interesting life."
1:59 p.m.
New subject: SparcServer 1000
On Sat, 31 Mar 2001, R. D. Davis wrote:
- One of the FRUs has what I'm guessing to be a
single older CPU
- The other FRU has what appears to be newer CPU modules, and these
Sounds like three SM51 modules to me.
Also found four 1.2 GB hard drives, two of which will
be used up by
Solaris.
1.2 GB.. those would be ST31200N drives, then? Bad news, if you've got
drives to replace them, I'd say do it. Every time we power off one of our
few remaining SSA110s (which are full of those disks), we have about a 50%
chance that one or more disks will never spin back up again.
I HATE those disks.
Alas, only three CPUs it seems.
Still not too bad.
Yeeeehaw!!! I've got the 32MB variety! :-) Well,
now it appears that
this machine has at least a fair amount of memory, as there are 12 of
the 32MB modules on each FRU, which amounts to a total of 768 MB of
total RAM... with that much RAM, is it still standard to use double
the amount of RAM for swap as a rule of thumb?
Yes, usually machines with that much RAM are solving problems which
require lots and lots and lots of memory. (:
On the Suns I manage, we do double physical RAM unless some beancounter
horribly overspecified the system on the PO. Never do less than 125%
physical, or else you can run into trouble with savecore if the machine
ever crashes. The biggest system I've ever managed had 48 GB RAM and 250
GB or so of swap---and was appropriately sized for the task. That was in
the data warehouse at Amazon.com. I think my NDA is up by now and they've
probably changed things enough that I can talk about it.
ok
r.
1 Apr
1 Apr
2:39 p.m.
New subject: SparcServer 1000
501-2541 - motherboard, 0 MB, (2) SM51, [SS1000]
501-2336 - motherboard, 0 MB, FRU, [SS1000]
http://www.sunstuff.org/hardware/partnumbers/501/2/3/3/6/
The 2336 is a Field Replacement Unit, no specs on processors on it, the
2541 is a dual SM 51 (55 ish Mhz proc) board.
The SBUS is at 20 Mhz.
You will need to look at the Barcode #'s on the MBUS CPU modules to tell
what speed they are, the barcode is -USUALLY- on the MBUS connector so you
mau have to pull the modules from the board to find out. I would expect
that the modules on both of your boards are SM 51's.
Information is part from memory, part from http://www.sunstuff.org
http://www.sunstuff.org/hardware/partnumbers/motherboards.shtml
At 01:43 PM 3/31/01 -0500, you wrote:
Greetings,
Is anyone here familiar with the Sun SparcServer 1000 systems? This
morning, I picked one up at a hamfest; it has two CPU boards in it,
each, it appears, has 4 CPUs. I was able to identify one of the
boards from it's part number from what I found on the net; appears to
have 20MHz CPUs; the other board is a higher number, but I've found
little data for it this far. The CPU boards are: 501-2336 and
501-2541. Does anone have any data on these?
Being unfmamiliar with multiprocessor systems, I'm wondering how fast
this will be in comparison to my 170MHz Ultra-1; slower for some
things and faster for others, I guess, while consuming much more
power---is this correct?
Wow, this is one heavy machine! In order to figure out if Solaris 8
can run on it, do I count the total amount of memory in the system, or
the amount of memory per CPU? Haven't identified the memory in it
yet, but one CPU board has 12 banks filled.
--
Copyright (C) 2001 R. D. Davis Ancient guru who see future once say:
All Rights Reserved "wise man who meditate upon immortal wisdom
rdd(a)perqlogic.com 410-744-4900 of Bart Simpson's words: "Don't have a
cow,
http://www.perqlogic.com/rdd man" will enjoy interesting life."
2 Apr
2 Apr
5:53 a.m.
New subject: SparcServer 1000
Being unfmamiliar with multiprocessor systems, I'm
wondering how fast
this will be in comparison to my 170MHz Ultra-1; slower for some
things and faster for others, I guess, while consuming much more
power---is this correct?
With 3 CPUs your 1000 will probably give you close to the floating point
performance of your Ultra 1 on a typical cruncing application
(13 MB/processor) well distributed among the CPUs (no memory contention).
I don't know about integer performance because I don't deal too much with it.
The I/O busses in your Ultra-1 are likely to be faster in Ultra-1 than in the
1000. Depending upon the number and vintage of the disk drives in each this
may or may not be a constraint.
The 4 processor 1000 I have access to here clocks in a 0.00053 second
Ahl's simple benchmark (4 process multiprocessing). The Ultra-10
(350MHz) on my desk does about 0.00078 seconds. Of course that's a 7k
executable, so it caches pretty well, but the 1000 beats the Ultra in
raw CPU. The 1000 has a non-Sun CPU upgrade, so its times should be better
than a stock 1000.
Of course the primary difference will be memory bandwidth. The Ultra-10
can move 123 MB/s (stream_d) while 1000 is lucky to get 32 MB/s.
Eric
31 Mar
31 Mar
9:33 p.m.
New subject: [pups] Reformat RA81/82?
On Sat, 31 Mar 2001, Gregory R. Travis wrote:
Can RA81/82 disks be reformatted in the field? If so,
is this
done through XXDP or is there a way to do it from the disk's
serial port?
Yes, yes and no. If we're talking PDP-11s, you need XXDP.
My RA82 occasionally lights its FAULT light and
extinguishes its
ready light during write operations. The likelyhood of it
faulting appears (but I am not certain) to be correlated with a specific
region of the disk.
About five-ten seconds after the fault, the FAULT indicator turns
off and ready comes back on. At no time is an I/O error generated
that the application or the operating system (2.11BSD) see
so this fault appears to be transient and is resolved by either
the drive or the drive/controller together.
Hmmm. Normally, bad spots developing on the disk are reported by the
controller. The OS is supposed to remap those bad blocks (BSD don't do
this), so it should never be hidden by the disk.
It "smells" like a data write error
that's resolved after a few
automatic retries. I'm hoping a low-level reformat could clear it
up. Am I wacked out?
Not sure, but it smells fish...
If there's an XXDP exerciser/formatter available
I'd appreciate it
if someone could point me there. It's been nearly twenty years since
the last time I even tried running XXDP
CZUDE (UDA disk formatter)
CZUDC (UDA and disk drive diagnostics)
All according to an old manual I have. My memory tells me that this
information is obsolete, and that new diagnostics were created, but I
can't find that information...
Hmm, a KDA50 manuals says:
CZUDK (KDA50 Disk drive formatter)
CZUDH (UDA-50/KDA-50 Basic Subsystem Diagnostics)
CZUDI (Disk Exerciser test)
CZUDJ (Subsystem exerciser test)
CZUDL (Bad Block Replacement Utility)
CZUDM (Disk Resident Error Log Utility)
I think these all are the current diagnostics for the UDA-50 also.
Johnny
Johnny Billquist || "I'm on a bus
|| on a psychedelic trip
email: bqt(a)update.uu.se || Reading murder books
pdp is alive! || tryin' to stay hip" - B. Idol
8446
days inactive
8448
days old
7 comments
6 participants
participants (6)
-
bqt＠update.uu.se -
geoffr＠zipcon.net -
greg＠ciswired.com -
korpela＠ellie.ssl.berkeley.edu -
rdd＠smart.net -
red＠bears.org